How the rankings are scored

This page states how a Kagit daily ranking score is computed, what each vertical actually measured, where the data came from, and when we say the data is insufficient. The scoring axes are not a fixed list — they are the axes that carried real data that day.

The formula

  1. 1. Renormalise the weights over the axes that were ACTUALLY measured. A missing axis is not a zero — scoring it as zero punished entities for being measured on more axes.
  2. 2. Shrink toward the category mean by how much was measured: λ = W ÷ (W + 0.25), where W is the weight mass of the measured axes. An entity riding one lucky axis is pulled back.
  3. 3. Gate before ranking: weight mass W ≥ 0.6, at least 2 measured axes, and evidence from at least 2 distinct providers.
  4. 4. Smooth with a 7-day exponential moving average so one noisy day does not throw the order.

Why there is a gate

An entity measured on one or two axes can post a high score, but that reads 'this is all we happened to find', not 'this is popular'. The gate keeps those rows out of the ranked order instead of letting them sit in it.

What the “insufficient data” badge means

Insufficient dataThe row did not clear step 3. It is not removed from the page — it sits in a collapsed section below the ranking, because hiding it would mislead more than labelling it. It says our coverage is thin, not that the entity is unpopular.

What each vertical actually measured

Read off each vertical's most recent validated run, not off a design weight table. An axis with no data stops appearing here.

VerticalScoring axes (weight)SourcesAs of
K-Pop Buzz Chart
  • Domestic Buzz (43%)
  • Consensus (22%)
  • Engagement (21%)
  • Chart Power (14%)
  • not counted this run: Global Reach
Bugs, Circle Chart, Google Trends, Melon, NatePann, Spotify, TheQoo, Wikipedia, YouTube2026-09-08
K-Beauty Ranking
  • Platform Ranking (37%)
  • Global Reach (25%)
  • Review Quality (25%)
  • Consensus (13%)
  • not counted this run: Community Buzz
GlowPick (글로우픽), Hwahae (화해), Musinsa Beauty (무신사 뷰티), OliveYoung (올리브영), YouTube Beauty2026-09-07
K-Drama Chart
  • Domestic Buzz (25%)
  • Global Reach (25%)
  • Streaming Rank (25%)
  • Engagement (15%)
  • Consensus (10%)
NatePann Drama, Netflix Global Top 10, Netflix Korea Top 10, TheQoo Drama, Wikipedia Pageviews2026-09-07
K-Entertainment Chart
  • Domestic Buzz (35%)
  • Global Reach (30%)
  • Engagement (20%)
  • Consensus (15%)
——
K-Food Trends
  • Popularity (45%)
  • Global Reach (33%)
  • Consensus (22%)
  • not counted this run: Engagement, Domestic Buzz
YouTube K-Food, 다이닝코드, 식신2026-09-07
K-Style Chart
  • Market (40%)
  • Global Reach (33%)
  • Consensus (14%)
  • Engagement (13%)
  • not counted this run: Domestic Buzz, Resale Premium
Musinsa Fashion (무신사), YouTube K-Style2026-09-07
K-Influencer Index
  • Audience Engagement (25%)
  • Content Authority (25%)
  • Platform Reach (20%)
  • Category Consistency (15%)
  • Community Trust (15%)
—2026-09-07

Change log

  • Switched to renormalising over measured axes, and added coverage shrinkage plus the ranking gate. Missing axes had been scored as zero, so an entity measured on more axes ranked worse.

← Rankings

Weights are renormalised over the axes actually measured, the result is shrunk toward the category mean by measured coverage (lambda = W / (W + 0.25)), rows below the coverage gate are held out of the ranking, and the series is smoothed with a 7-day EWMA.